How to Measure Preference for AI Engines

Beyond Presence: The Quality of Your Mention

When ChatGPT lists five CRM platforms, the order matters. The framing matters. Whether it says "leading platform" versus "one option among many" makes a material difference to the reader who never clicks through to verify. AI answers carry implicit authority, and the positioning within that answer shapes buying decisions.

Preference is the second dimension of the AVI Score. It measures how favorably AI systems describe your brand compared to competitors. A high Presence score with low Preference means you appear often but are consistently positioned as a secondary or inferior option.

The Three Layers of Preference

We break Preference into three measurable layers. Each contributes to the overall sentiment score.

• Positioning: Where in the answer does your brand appear? First mention carries disproportionate weight. Being listed as "the top choice" versus "also worth considering" signals different levels of endorsement.
• Sentiment: What language does the AI use? We analyze adjective patterns, comparative framing, and qualifier words. "Industry-leading" versus "budget-friendly" versus "controversial" paint very different pictures.
• Recommendation strength: Does the AI explicitly recommend your brand for specific use cases, or does it present all options neutrally? Some brands trigger strong recommendation language ("ideal for enterprise teams") while competitors get generic descriptions.

What Shapes AI Preference

AI models form preferences from patterns in their training data and retrieval sources. If the majority of web content about your brand is positive, expert-driven, and contextually rich, AI will reflect that. If the discourse is mixed or dominated by competitor comparisons where you lose, AI will reflect that too.

• Review sentiment across platforms (G2, Capterra, Trustpilot, Google Reviews) directly influences how AI frames your brand
• Expert endorsements and industry analyst coverage create authoritative signals that AI models weight heavily
• Comparative content where your brand wins (vs. pages where competitors are positioned as superior) shapes the recommendation pattern
• Your own content tone and positioning: if your website positions you as "affordable" rather than "premium," AI will follow that framing

Measuring Preference in Practice

For each query in our test set, we analyze the AI response along three axes: where your brand appears in the answer (position), what language is used to describe it (sentiment), and whether it receives an explicit recommendation (strength). Each axis scores 0-100, and they combine into an overall Preference score.

We also track Preference Drift, which measures whether your sentiment is improving or declining across monthly measurements. A brand can have a decent Preference score today but a negative drift, signaling that competitor activity or new training data is eroding its position.

Improving Preference

Unlike Presence (which can often be fixed with technical changes), improving Preference requires strategic content and reputation work. You need to shape the narrative around your brand across the web, not just on your own site.

The most effective lever is third-party content. Encouraging satisfied customers to leave detailed reviews, publishing case studies with specific metrics, and securing coverage in industry publications all feed the signals that AI uses to form positive preferences. On-site, ensure your positioning language is deliberate. The words you use to describe your own products become the words AI uses.